2022 Undergraduate Research Showcase

Document Type

Student Presentation

Presentation Date

4-22-2022

Faculty Sponsor

Dr. Elisa Barney Smith

Abstract

Sheet music is easier arranged, stored, and analyzed in a digital format. Machine learning algorithms can be used to recognize the content of printed sheet music and convert them to MusicXML digital notation. Object detectors, like YOLOv5 and RCNN are used to create a music symbol alphabet, starting with staff lines. These object detectors were then compared to each other for classification effectiveness in detecting the first symbol in the music alphabet, staff lines. Initial results suggest the applied approach for detection is effective, with YOLOv5 having higher accuracy and efficiency of detection when compared to RCNN. Further detection will be done to extract more symbols such as clef marks and music notes using the same approach.

Share

COinS